Avian influenza A/H5N1 was first identified in the late 1990s. To date, it has caused outbreaks in poultry or wild birds in 58 countries resulting In at least 571 human cases, and 335 human deaths, and more than $10 billion in economic losses, largely due to culling of poultry. Despite radical, often draconian, control efforts. H5N1 persists across many parts of the Old World and continues to threaten global health and development. In this proposed work we will use a combination of three different modeling approaches, collection of field surveillance data on domestic poultry and wild birds and a contact survey in /VH5N1 endemic countries to investigate what factors lead to the endemic persistence of A/H5N1. We will integrate dynamical, spatial and network models to develop a predictive capacity for the spillover of H5N1 from poultry to people. We have assembled a multidisciplinary team of leading disease ecologists, leveraging many years of experience in three H5N1 endemic (Bangladesh, China, and Egypt) and one control country (Cameroon, where H5N1 is not endemic). We will collect data on wild waterfowl migration, poultry farm size, market dynamics and human contact networks with these sources of infection. We will collect samples and examine data on viral incidence within animal populations at 10 sites in each country. We will test 3 hypotheses to understand why H5N1 is able to persist endemically in these countries: (1) The pattern, size and distribution of poultry farms drives H5N1 endemism; (2) The pattern, frequency and intensity of contact networks among wild birds, domestic birds and people influences the risk of spillover and prolongs transmission; and (3) Integrating dynamical, network and spatial models across scales provides a strategy to better predict influenza-A transmission risk. This work will significantly advance our understanding of the long-term dynamics of H5N1 by using fine-scale measurements of realistic contact networks in epidemiological models of H5N1, allowing us to explain the poorly understood capacity of H5N1 to persist at low prevalence in endemic countries. Furthermore, the improved predictive models will aid in developing more effective control measures in these critical, high risk countries.